Closed RunpengLuo closed 5 months ago
Hi Runpeng,
In fact, I never knew that Clair3 runs on Mac via docker. I just played with it, and I found that it was the docker virtualization framework that enabled such an easy deployment of Clair3 on Mac.
That is to say, the docker runs upon a virtual machine, which explains why it runs slower. To run Clair3 natively on Mac, here is a document. It is complicated, and the steps need a bit of tuning in each system I believe. I just tried the steps in the document on my Macbook M2 Max with Sonoma 14.5 and updated the document to get all the steps running again. Running natively on Mac means that Clair3 will use GPU thus runs as fast or faster than running on CPU on a Linux server. If you use Clair3 frequently on Mac, it is worth some effort to set it up.
Hi Dr. Ruibang,
I just try it on the linux server and it worked smoothly and finished in several minutes. Really appreciate your time on exploring this. For now I will stick with linux server but happy to try the Mac method via the linked document as well. Thanks again for your effort!
Hello,
Thanks a lot for developing this great variant caller Clair3. But I'm facing an issue when calling the variants against a small 2MB bacteria genome.
In particular, I have ONT read data with 40MB base called via Dorado sup mode. I pulled docker image via
docker pull hkubal/clair3
, and I used the modelr1041_e82_400bps_sup_v410
from https://github.com/nanoporetech/rerio/tree/master/clair3_models for Clair3. I used the following command for Clair3.I've got the following logs.
It took nearly 10 hrs to finish the whole execution under MacOS M2 chip, and I can see that the docker is running in full demand but doesn't seem to be paralleled throughout the execution (CPU utilisation is ~100%). Even when I downsample the reads to 1x, it still takes ~1.5hrs to finish the pileup step and stuck at the calling step. Just wondering whether this is normal and if there is any parameters I can use to speed it up, thanks a lot for your help!
John